Each team is assigned 5 secret target words by the Judge. Teams must craft a prompt that subtly leads ChatGPT to talk around the opponent's target words. No root word or derivatives of the target term may be used. All prompts are executed using the designated ChatGPT device provided. The goal is to: Make ChatGPT respond relevantly, without being too obvious. Prevent the opponent from correctly guessing your assigned word. In the guessing phase, the teams will be handed a word bank with a100 words from which they would have to find the answers to the 5 riddles given. If they guess the wrong word, the points will be given to the team which the word (wrongly guessed) actually belongs to. The prompt should be interpretable by the LLM but misleading enough that the opponent cannot guess the target word.

• Teams can join anytime during this round.
• Flexible entries: Any team arriving during this window can participate.
• Each team plays one match, facing another team.
• Judges maintain scores for each team based on a scoring rubric.
• Top 4 teams with the highest cumulative scores progress to the Final Rounds.
• Time Limits:
• 10 minutes for prompt crafting
• 5 minutes for the opponent's guessing

The scores for each of the teams are evaluated on the basis of the criteria decided, which includes Prompt Quality, LLM Interpretability, Deceptiveness, Opponent's guess, Clarity and Style, in case of a Tie, the team with higher LLM Interpretability score wins.

In the Final Rounds, each team must incorporate a randomly drawn "Power Prompt Twist Card" into their crafted prompt. Examples of twists: Use a pun in your prompt Include a metaphor involving space Prompt must be exactly 25 words Prompt must be phrased as a question Prompt must rhyme This twist increases difficulty and forces teams to think creatively. Successfully integrating the twist without compromising deception or clarity is now a part of the judging criteria.

• Semifinal 1: Top Seed vs 4th Seed
• Semifinal 2: 2nd Seed vs 3rd Seed
• Grand Finale: Winners of both semifinals face off
• Failing to adhere to the condition in Power Prompt Twist card impacts scoring under "Prompt Quality".

The scores for each of the teams are evaluated on the basis of the criteria decided, which includes Prompt Quality, LLM Interpretability, Deceptiveness, Opponent's guess, Clarity and Style, in case of a Tie, the team with higher LLM Interpretability score wins.